Introduction
Do schools in Panama get better access to public internet? This research will analyze whether schools in Panama, get access to public internet via access points. We will also look into, the spatial correlation of access points, where we want to highlight if access points are either located in high density areas or not. As a plus, I will also look into the spatial relationship of access points and schools with the Panamerican Highway, which is the main highway that crosses the country. For this project we will be using the following data sources:
- Schools locations: Ministry of Education of Panama
- Access Points locations: Governement Innovation Institution
- Panamerican Highway location: Smithsonian Tropical Research Institution
Hypothesis
For this project I have the following First-Order and Second-Order properties:
- First-Order Property: I will analyze how the intensity of access points varies across Panama in relation to schools and population density. My hypothesis is that the density of access points is higher in areas with higher population density and greater concentration of schools, suggesting that infrastructure deployment follows population and educational needs.
- Second-Order Property: For my second-order property, I want to explore the spatial relationships between access points, schools, and population density. My hypothesis is that access points exhibit positive spatial autocorrelation with both schools and population density, indicating that their placement is influenced by these factors and tends to cluster in areas with higher educational and population demands.
Let’s begin below, with the Exploratory Data Analysis.
Exploratory Data Analysis (EDA)
Below we display two tables for both Districts and at the Province level showing relevant indicators such as: - Numbers of schools - Numbers of Access Points - Population - Access Points per School Ratio - Access Points per 1000 people Ratio
This is useful to compare, across the project, how more densed districts behave spatially.
We can see from the above tables, that Districts that house cities, are the ones with higher amounts of access points. We see the same behavior with the Provinces, specially with main cities such as Panama, Veraguas and Chiriqui.
Most of these studies, however, are based on observations of conflict events. In this study, we study the more fundamental variable of a capital’s distance from the population centroid of the country.
Correlations between Population, Schools and Access Points
Below I show 2 plots, where on the left we see the relationship between Access points and Population and on the right, we see the relationship between Access Point-School ratio with the Population.
Statistical Regression of Access Points on Population
I have run 2 statistical regression models, where I regress Access Points on Population and Access Points both on Population and Schools. Adding schools improves the model’s explanatory power (R² increased by 2.6%). Both population and schools are significant predictors as we can see in the below tables. We can see that for each additional school, we expect 0.216 more access points, holding population constant.
| Model 1 | Model 2 | |
|---|---|---|
| + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | ||
| Intercept | 4.874*** | -2.568 |
| (1.302) | (1.772) | |
| Population | 0.000*** | 0.000*** |
| (0.000) | (0.000) | |
| Schools | 0.216*** | |
| (0.040) | ||
| Num.Obs. | 75 | 75 |
| R2 | 0.909 | 0.935 |
| R2 Adj. | 0.907 | 0.933 |
Geospatial Analysis
Access Points Map
Let’s explore how access points are distributed along the country. As validated above, the districts with the highest amount of access points are those located around Panama City (459 approx), Santiago with 198 and Boquete with approximately 118 access points. Interestingly we can detect that probably these clusters, do not follow a random location. I have clustered the access points data, to facilitate the readibility of the map. We can see the first signs that the spatial distribution follows a “S” layout, same as the country. This will be more clear later on, when we display other points.
Access Points Density per District Map
Now let’s explore how different access points density is within districts. With this map we can confirm that the districts of David, Santiago, Colon and Panama are the ones with a higher concentration of access points. This makes sense in the context that, these districts are where we can locate higher economic development in the country.
Panama Highway Map
Highways represent economic development, as they try to bridge different disitrcts across the country. Our hypothesis here is that, access points will be located, around districts where we can see an intersection with the Panamerican Highway. It is worth noting that this highway goes across the complete country form west to east, mostly located on the pacific side of the country, where Panama City is.
Public Schools Map
This map shows, clustered, every school in the country. There are 9 types of classifications on the school system:
- COIF: Early Childhood Attention Center
- Cefacei: Early Chilhood Community and Familiy Centers
- IPHE: Special Habilitation Panamanian Institute
- Kinder: Pre Kinder
- Parvulario: Kinder
- Primaria Oficial: Elementary School
- Privada: Private School
- Secundaria: Highschool
- Universidad: College
Which Districts are Intersected by the Panamerican Highway?
Below we find, which are the districts that intersect with the Panamerican Highway. As confirmed by the above analysis, the districts with the higher population densities and with the higher amount of access points installed, are more likely to be intersected by it. This is a causal effect of infrastructure development, as highways are more likely to be equiped with better structures to implement different services such as telecommunications. As seen on the maps above, we can see that the districts with a higher concentration of access points, are the same ones that intersect with the highway.
Moran’s I
Access Points Moran’s I
We know that the scale runs from -1 (perfect dispersion) to +1 (perfect clustering). Our result is 0.229, which indicates positive spatial correlation meaning that districts with similar number of access points (higher or lower) tend to cluster together. With our p-statistic, we can reject the null hypothesis of random spatial distribution. This means that access points are not positioned randomly but rather logically.
[1] "Moran's I for Access Points vs Schools:"
Moran I test under randomisation
data: district_data_complete$access_points
weights: nb2listw(nb, style = "W", zero.policy = TRUE)
n reduced by no-neighbour observations
Moran I statistic standard deviate = 4.7929, p-value = 8.218e-07
alternative hypothesis: greater
sample estimates:
Moran I statistic Expectation Variance
0.229408902 -0.013888889 0.002576752 Mapping Access Points vs Schools ratio
As expected, we have a higher ratio within districts that are city districts such as Panama, San Miguelito, Pedasi and Chitre. It is worth mentioning that for instance, Pedasi is not a big city town, but instead is touristy town. They have 9 school total, with 11 AP located in this district. In more urban areas we will have a higher ratio of access points to schools. This signals to a possibility of schools being located in highly densed areas and therefore, finding more access points.
Pairwise Intensity function
Analyzing Intensity function
As expected, we see that access points tend to cluster together within distances of 1500 meters. When that threshold is passed, the clustering of these points decreases sharply. This is consonant to the logic of installing access points in dense areas and where more schools are located. The Mean nearest neighbor distance is 1617.98 meters. This means that we have some outliers access points, that are located in difficult access areas in very rural zones. They are more likely to be separeted from others. These locations could be pulling the mean up. On the other hand the Median nearest neighbor distance is 400.99 meters which is lower than the mean. This indicates that most of the access points are clustered in distances lower than 400 meters. This indicates a clustered pattern and not a uniform across the country. These are likely to be clustered in urban areas.
Mean nearest neighbor distance: 1617.98 meters
Median nearest neighbor distance: 400.99 metersBuffer Analysis with Panamerican Highway
As expected, from this analysis, we can see that there is a higher amount of schools outside the inner buffers, as schools will try to serve the wider population, beyond the urban areas. On the other hand, we will see more access points located in urban areas, therefore closer to the highway as they serve more the densed urban areas. Now if we analyze and see the amount of schools greater that 5km from the highway we see the number increaseing. This logically means that location of schools it is not as influenced by the highway as access points are. Schools follow a more uniformed pattern across the country, without clustering. This gives higher access to school to areas that are rural as well.